review (XS) for 6879943: CTW failure jdk6_18/hotspot/src/share/vm/c1/c1_LIR.hpp:2029

[Tom Rodriguez]

On Jan 15, 2010, at 11:27 AM, Changpeng Fang wrote:

> On 01/15/10 11:13, Tom Rodriguez wrote:
>> What do you mean?  It's not experimental at all.
>> 
>>   
>> 
> Why is maxNumberOfOperands 16 (not 18, ...)? Is it because
> 16 is the smallest number to solve the problem?

Yes.  It's derived from testing.  It's likely possible to derive it from a combination of other values or to make it more adaptive but that's more complicated.  Using a fixed size has worked ok so far and I just wanted to increase it for this issue.

tom


> 
> It looks to me that you change the IR specification (maxNumberOfOperands=16)
> based on experiment, which is totally fine with me.
> 
> Thanks,
> 
> Changpeng
> 
> 
>> tom
>> 
>> On Jan 15, 2010, at 11:05 AM, Changpeng Fang wrote:
>> 
>>   
>> 
>>> On 01/15/10 10:45, Tom Rodriguez wrote:
>>>     
>>> 
>>>> http://cr.openjdk.java.net/~never/6879943
>>>> 
>>>> 
>>>>   
>>>> 
>>>>       
>>>> 
>>> The fix itself is OK. But, is  maxNumberOfOperands 
>>> purely experimental?
>>> 
>>> Changpeng 
>>> 
>>>     
>>> 
>> 
>>   
>> 
>